Fine-grained Visual Reasoning on V* (Avg@8)
89.5Avg@8 OverallMAPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MAPObackbone=Ovis2.5-9B2026.04 | 89.5 | 91.1 | 87.2 | |
| Mini-o3reproduced_by_authors=true2026.04 | 89.3 | 90.4 | 88.2 | |
| Ovis2.5-9B + GSPOtraining=GSPO2026.04 | 89.1 | 87.1 | 92 | |
| Ovis2.5-9B + DAPOtraining=DAPO2026.04 | 88.8 | 88.1 | 90 | |
| Ovis2.5-9B + GRPOtraining=GRPO2026.04 | 88.6 | 88 | 89.5 | |
| DeepEyesreproduced_by_authors=true2026.04 | 88 | 89.6 | 85.5 | |
| Ovis2.5-9B + PPOtraining=PPO2026.04 | 82.8 | 79.1 | 88.2 | |
| Thymeresults_source=original paper2026.04 | 82.2 | 83.5 | 80.3 | |
| DeepEyesv2results_source=original paper2026.04 | 81.8 | — | — | |
| Ovis2.5-9B2026.04 | 81.8 | 81.9 | 81.6 | |
| Gemini 2.5 Pro2026.04 | 81.4 | 85.5 | 75.2 | |
| Ovis2.5-9B + Coldstart SFTtraining=Supervised Fine-Tuning2026.04 | 77.5 | 73 | 84.2 | |
| GPT-52026.04 | 71.3 | 70 | 73.4 |